Physicians’ prescribing behaviors can be categorized as rational, semi-rational, or irrational (see Fig. 1). We hypothesized that a irrational prescribing behavior dominates in the TOP10 (Bindel and Seifert 2024b). The classification of rational, semi-rational, and irrational presription behavior is characterized by contrasting relationships between the development of bacterial resistance and DDD-prescriptions. A negative correlation between these variables suggests rational or semi-rational prescribing behavior, as both approaches consider bacterial resistance in decision-making. In rational prescribing, an increase in bacterial resistance would lead to the exclusion of the respective antibacterial drug, as treatment decisions are guided by specific pathogen information, resulting in a strong feedback loop. Semi-rational prescribing also takes resistance into account, though the feedback may be less immediate due to reliance on broader data sources and delayed analysis. Conversely, a positive correlation indicates irrational prescribing behavior, where bacterial resistance is not a factor in the selection process, and decisions are primarily driven by cost considerations, leading to the continued use of antibacterial drugs with increasing bacterial resistance.

The correlation between bacterial resistance and DDD-prescriptions can be initially characterized by examining the number of significant and strong correlations. Given the wide variability in the number of pathogens and thus the number of possible correlations, comparability can be achieved by calculating proportions. The strength of influence is measured by the share of strong versus total correlations, while the range of influences is determined by the proportion of significant versus total correlations. Tables 1 and 2 summarize the analyzed aspects.

Doxycycline, clindamycin, and clarithromycin exhibit the highest share of significant correlations among total pathogens at 66.7%. Following at a considerable interval are amoxicillin clavulanic acid (44.4%), ciprofloxacin (42.9%), and amoxicillin (40.0%). Cefuroxime axetil (28.6%) and nitrofurantoin (20.0%) show relatively low shares. Azithromycin demonstrates a share of 0.0% due to the non-significance of all analyzed pathogens. A larger influence of bacterial resistance on DDD-prescriptions corresponds to a greater number of significant correlations among the analyzed pathogens. Conversely, a lower proportion suggests a minor influence of bacterial resistance on DDD-prescriptions for most pathogens.

When examining the proportion of strong correlations among total correlations, ciprofloxacin shows the highest share at 28.6%, followed by amoxicillin clavulanic acid and sulfamethoxazole-trimethoprim at 11.1%. Other antibacterial drugs do not exhibit strong correlations, resulting in a share of 0.0%. A higher share indicates a stronger influence of bacterial resistance on DDD-prescriptions, with antibacterial drugs affected by multiple pathogens. A lower share implies limited influence by few pathogens, while a zero share indicates no strong influence by any pathogen.

There is little indication for the presence of rational prescribing behavior. Only clindamycin exhibits consistently negative correlations (see Table S5), although these may be a result of a distortion in DDD-prescriptions in 2012 (Bindel and Seifert 2024a). A comparison of the developments of bacterial resistance and DDD-prescriptions suggests that the real correlations might be positive. For azithromycin, no assessment is possible as all correlations are non-significant (see Table S6).

Semi-rational prescribing behavior is apparent for amoxicillin clavulanic acid, since both positive and negative correlations are observed in a balanced ratio (see Tables 2 and 4). Sulfamethoxazole-Trimethoprim also have both positive and negative correlations including a strong negative correlation. These data point to semirational prescribing of sulfamethoxazole-trimethoprim.

For most prescribed antibacterial drugs, a irrational prescribing behavior is present. This includes amoxicilline, cefuroxime axetil, doxycycline, nitrofurantoin, ciprofloxacin, and clarithromycin, since they only exhibits positive correlations (see Tables 2, 3, 4, S8, S9, and S10). For no antibacterial drug, we obtained evidence for rational drug prescribing.

To categorize the results and assess the strength of bacterial resistance’s influence on DDD-prescriptions, it is necessary to compare these results with other influential factors. DDD-costs are particularly suitable for this comparison, as they have been demonstrated to significantly impact DDD-prescriptions (Bindel and Seifert 2024b). By juxtaposing these factors, we can better understand the relative contributions of economic and microbiological considerations to prescribing behaviors for antibacterial drugs.

To facilitate a meaningful comparison, we calculated the mean value for each antibacterial drug. Given that both analyses are based on the same prescription data, this comparison is appropriate. Correlation with the longest available period was employed to minimize the effect of bias or short-term fluctuations. Table 4 presents the mean values obtained from the subsequent analysis.

The number of significant results is in both cases nine out of ten. However, the specific antibacterial drugs showing significant values differ between the two parameters. Notably, no significant average value for the correlations of bacterial resistances was observed for azithromycin, whereas doxycycline lacked significant values concerning DDD-costs (Bindel and Seifert 2024b).

The direction of correlation for each parameter also exhibits distinct patterns. Bacterial resistances predominantly show a positive correlation, with six positive and two negative correlations. Conversely, all correlations related to DDD-costs are negative.

Considering the number of strong correlations, defined by a correlation coefficient exceeding (+ / −) 0.8, there are notable differences. For bacterial resistance, ciprofloxacin is the only antibacterial drug showing a strong average correlation (see Table 1). In contrast, DDD-costs demonstrate four strong correlations, including amoxicillin, cefuroxime axetil, clindamycin, and nitrofurantoin (see Table 4). While the strong correlation is negative in DDD-costs, it is positive in bacterial resistances.

When comparing the magnitude of the parameters for each antibacterial drug, significant differences are apparent. In most cases, one parameter exhibits a strong influence while the other remains relatively minor (see Table 4). Typically, the abolute value of DDD-costs exceeds that of bacterial resistance. This includes amoxicillin, cefuroxime axetil, amoxicillin-clavulanic acid, clindamycin, sulfamethoxazole-trimethoprim, and nitrofurantoin. Conversely, for ciprofloxacin and clarithromycin, bacterial resistance has a greater impact. Due to the lack of one parameter, no assessment could be made for doxycycline and azithromycin. Summing the total of all significant values in each category reveals that DDD-prescriptions influenced by DDD-costs (− 0.743) exceed those influenced by bacterial resistances (0.254).

The reliability of the values for both parameters is high, given the availability of significant values for nearly all antibacterial drugs. Consequently, both factors can be considered stable and meaningful, allowing to draw conclusions. It is noteworthy that these parameters affect DDD-prescriptions in opposite directions. While bacterial resistance tends to develop in a manner similar to DDD-prescriptions, DDD-costs exhibit an inverse relationship. During the analyzed period, declining DDD-costs led to an increase in DDD-prescriptions, whereas bacterial resistance and DDD-prescriptions generally rose or declined similarly.

Thus, DDD-costs have a more substantial influence on DDD-prescriptions in outpatient settings than bacterial resistance in most cases. This suggests that the prescription of antibacterial drugs is primarily driven by economic considerations rather than rational and scientific facts. Although this prescribing behavior reduces health system costs, it undermines the efficacy of antibacterial drug use and promotes the development of bacterial resistance. Given the extended analysis period, this phenomenon has persisted for decades and is not merely a recent development.

The analysis in this study is based on data from the Arzneiverordnungsreport. As only outpatient prescriptions of the GKV system are included, no assessment can be made regarding prescriptions in hospitals or via private health insurance. As the data are based on developments in Germany, they are not directly transferable to other countries. No differentiation was made according to age or region.

Data collection for DDD-prescriptions and -costs depended on the design of the chapter under consideration. Changes in the structure of the sections over the years led to a bias for clindamycin in 2012 (Schwabe and Paffrath 2013). The ARS criteria for analyzing only indicated pathogens could exclude strong correlations, as a high increase in bacterial resistance might lead to the exclusion of the antibacterial drug in question. Both factors could lead to an under- or overestimation of the correlations in question.

Generalizations are restricted by the limited number of examined antibacterial drugs, pathogens, and factors considered. Data on bacterial resistance are only available for a short period of time compared to the available data on DDD-prescriptions and -costs. Developments in bacterial resistance prior to 2008 was not analyzed, which may lead to less accurate correlations. There may be other factors influencing the outcomes studied that were not included in the analysis.

Certain criteria were set for the statistical methods used. Between the choice of statistical procedures, the value for a significant correlation with a 2-sided significant level of 0.01 and 0.05, as well as a determined strong correlation coefficient above (+ / −) 0.8, was determined. Changing these parameters may lead to different conclusions.